This home humanoid costs $3,555 and learns new tasks with your smartphone
Flourish Robots opened pre-orders Monday for Flourish 1: a 1.1-meter, 20-kilogram wheeled humanoid that sells for $3,555 and learns household tasks from a smartphone demo.
First deliveries are scheduled for December.
The price is real. Not a concept, not an industrial unit with a consumer sticker: a 50-unit pre-order run for actual homes, at a price point that puts a general-purpose home robot in reach of consumers for the first time.
How the 30-minute learning actually works
This is the interesting part ...
You teach Flourish 1 a task by holding your phone and moving it naturally: picking up shoes, wiping a table, loading a dishwasher tray. The phone's built-in inertial measurement units (IMUs) capture the motion data. That data goes to cloud-based GPU compute, which trains a skill model based on your demonstration. About 30 minutes later, the robot can replicate the task using its two arms.
No coding. No teleoperation rig. No robotics expertise required.
(At least, that's the promise. We'll believe it when we see it.)
Users can optionally run the training on their own computers instead of the cloud. Wrist-mounted cameras on the grippers handle the manipulation tasks; a small lidar unit near the base handles obstacle detection.
The hardware choices
The robot runs on Raspberry Pi.
Founder Antoine Marcel made no attempt to obscure this: cost is the point. Raspberry Pi is inexpensive, widely available, and well-understood by a global developer community. It's not the compute you'd choose if your goal were maximum robot intelligence ... it's the compute you choose if your goal is a product that can retail at $3,555 and still make economic sense to manufacture.
Six wheels — not legs — provide stability without the complexity of a bipedal system. The design is deliberate: wheels are reliable, quiet, and cheap.
The arms and manipulation capabilities are where the differentiation happens, and each arm can carry 1.5 kilograms.
What it can and can't do
Flourish 1 can be trained to pick up shoes, clothes, and toys.
It can load a dishwasher, wipe tables, water plants, and fetch items. Apparently it can also clean a cat litter box (impressive if accurate). It can't do dishes because it's not waterproof, and can't perform intricate assembly work.
The current generation is focused on tidying and object handling rather than anything requiring fine dexterity.
That constraint set is honest and probably correct for a first product at this price.
The question for a wheeled robot running on a Raspberry Pi isn't whether it can do everything ... it's whether it can do something useful reliably enough that people keep using it.
The 12-hour battery life suggests Marcel is thinking about all-day deployment, not episodic demos.
Why home robots keep failing, and what's different here
The history of home robots is mostly a history of narrow products that couldn't adapt.
Roomba worked because it did one thing (vacuum flat floors) and did it well enough. Every attempt to build a more general home robot has foundered on the same challenge: every household is different, with different layouts, different objects, different routines.
Pre-programmed behaviors don't cover that ... you need robots that can learn.
The smartphone IMU teaching mechanism, if it works at the quality the 30-minute claim implies, is a genuine attempt at a different answer. Instead of pre-programming or teleoperation, you demonstrate. Instead of hiring a robotics engineer, you pick up your phone. The learning loop runs in the cloud and comes back as a skill the robot can execute.
The first 50 units will tell us whether the claim holds up in real households. That's a small number, but it's enough to know.
Sources
Written by
I tell stories with data and build startups. I've raised angel and VC funds, led software companies and startups, and chronicled the rise of the mobile economy. Currently I write for Forbes, host several podcasts, and am starting up HumanoidMegaHub. I was TUNE's mobile economist and Singular's VP of Insights, built the research team at VentureBeat, and managed teams creating software for partners like Intel and Disney. In past lives I built websites, e-learning platforms, and mobile apps, and, a long time ago, was named to Folio's top 100 of the media industry's "most innovative entrepreneurs and market shaker-uppers.”